Hedwig Nora Nordlinder
The standard administrative subdivision of Sweden is Country -> County -> Municipality. All of these have legal significance and their borders have historically been determined by political decisions. Alongside the standard administrative subdivisions Statistics Sweden (Statistiska Centralbyrån) as well as other government agencies have created subdivisions for statistical purposes, such as the Regional Statistical Areas (RegSO), Demographic Statistical Areas (DeSO) and local labour markets (lokala arbetsmarknader). I will in this post take a special interest in the local labour market subdivision. The material is quite technical (admittedly more than initially intended) so I envision the primary target audience to be the people responsible for developing and maintaining the local labour market standard in various countries.
The local labour markets are created by first determining which municipalities can count as “local centers”. A local center is a municipality in which
at least 80% of the population works in it
at most 7,5% of the population commutes to any other specific municipality
After the local centers are created all remaining municipalities are assigned to the labour market created by the local center that most of its commuting residents commute to. (Construction and use of labour market areas in Sweden, Statistics Sweden 2010).
Quoting directly from the aforementioned SCB report: “The purpose of local labour markets is to describe the functioning of the labour market for geographical areas that are relatively independent of the outside world with respect to supply and demand of labour” (translation my own, original source in Swedish).
This post will ask and answer the question: Does this method of subdivision work by testing whether or not the local labour markets are mostly independent of the outside world. To perform this test this I will analyse how this subdivision type interacted with the COVID-19 pandemic.
In 2020 the COVID pandemic swept over the world, with Sweden being no exception. It is perhaps the best monitored and documented pandemic in the history of the world. In particular, the granular data collected by the Swedish Public Health Agency (Folkhälsomyndigheten) provides an excellent opportunity to test if the labour market regions live up to their purpose of being “relatively independent of the outside world”
If labour market regions are truly independent economic units we would expected COVID-19 spread patterns between adjacent labour markets to be less correlated than say, between counties (whose borders are not drawn with the explicit purpouse of defining self contained economic zones)
This formulation will lead the layman to ask questions such as “What counts as ‘less correlated’” and the expert asking questions such as “How do we define ‘correlation’”. I will make explicit my modeling assumptions later in the post, but first, let us visualise the spread of COVID through Swedish municipalities. While I wish to keep the amount of code displayed to a minimum, credit is due to Filip Wästberg for the creation of the R package “swemaps2”, which I make heavy use of.
Now that we have installed swemaps2 we can make a map of Swedens labour market regions. They changed ever so slightly during the pandemic, so we will have to account for that in our modeling.